The results of selective magnetic resonance imaging of the knee were compared with those of arthroscopy in a prospective series of fifty patients. A specifically designed protocol for imaging, producing T1 sagittal images interleaved at four millimeters while the patient's foot was in 20 degrees of external rotation, was utilized. This technique, called selective magnetic resonance imaging, yielded excellent visualization of the posterior cruciate ligament, medial meniscus, and lateral meniscus in all patients. However, in only 76 per cent of the patients was the anterior cruciate ligament well visualized. Compared with arthroscopy, the sensitivity, specificity, and accuracy of selective magnetic resonance imaging were, respectively, 95.8, 100, and 98 per cent for tears of the medial meniscus; 66.7, 95.1, and 90 per cent for tears of the lateral meniscus; undefined, 100, and 100 per cent for tears of the posterior cruciate ligament; and 100, 96.9, and 97.3 per cent for tears of the anterior cruciate ligament, when that ligament was well visualized. Our selective sequence can be performed in fifteen minutes at a cost that is comparable with that of arthrography. It is totally non-invasive and requires no exposure to ionizing radiation. Selective magnetic resonance imaging can be a safe and valuable adjunct to the clinical evaluation of the knee and an aid to efficient preoperative planning.